Exploratory Data Analysis

Mismatches in FDI data:

See here for an explanation on on the meaning of these mismatches.

df <- right_join(target_info, source_info,
           by = c("source" = "source", 
                  "target" = "target", 
                  "year" = "year")) %>% 
  mutate(diff = abs(flow.x - flow.y)) %>% 
  mutate(dyad = str_c(source, " -> ", target)) 

df %>%   
ggplot(aes(flow.x / 100, flow.y / 100, color = diff / 100)) + 
  geom_point(alpha = 0.6) + 
  geom_abline(slope = 1, intercept = 0, linetype = "dashed") +
  geom_hline(yintercept = 0, alpha = 0.5, size = 0.5) +
  scale_color_viridis_c(direction = -1, labels = scales::dollar) +
  scale_y_continuous(labels = scales::dollar) +
  scale_x_continuous(labels = scales::dollar) +
  labs(x = "target info", y = "source info", caption = "(billions of US dollars)", color = "mismatch")

Top offenders by year:

for (i in 2001:2012) {
  print(df %>% 
  filter(year == i) %>% 
  group_by(year) %>% 
  filter(rank(-diff) <= 10) %>% 
  ggplot(aes(x = fct_reorder(dyad, diff), y = diff / 100)) +
  geom_point() + 
  facet_wrap(~ year, scales = "free") +
  coord_flip() +
  scale_y_continuous(labels = scales::dollar) +
  labs(x = NULL, y = "Size of mismatch", caption = "(billions of US dollars)")
  )
}

Top offenders by year (weighted by target country’s GDP)

wdi_varlist <- readRDS("wdi_data/wdi_varlist.RDS")

gdp <- readRDS("wdi_data/wdi_data.RDS") %>% 
  select(country, continent, year, NY.GDP.MKTP.CD) %>% ## GDP (current US$)
  rename(gdp = NY.GDP.MKTP.CD) %>% 
  ungroup()

target_info <- target_info %>%
  left_join(rename(gdp, source = country)) %>% 
  rename(source_gdp = gdp, source_continent = continent) %>% 
  left_join(rename(gdp, target = country)) %>% 
  rename(target_gdp = gdp, target_continent = continent)

## Joining, by = c("source", "year")

## Joining, by = c("target", "year")

source_info <- source_info %>%
  left_join(rename(gdp, source = country)) %>% 
  rename(source_gdp = gdp, source_continent = continent) %>% 
  left_join(rename(gdp, target = country)) %>% 
  rename(target_gdp = gdp, target_continent = continent)

## Joining, by = c("source", "year")
## Joining, by = c("target", "year")

df <- right_join(target_info, source_info,
           by = c("source" = "source", 
                  "target" = "target", 
                  "year" = "year",
                  "target_continent" = "target_continent",
                  "source_continent" = "source_continent",
                  "source_gdp" = "source_gdp",
                  "target_gdp" = "target_gdp")) %>% 
  mutate(diff = abs(flow.x - flow.y)) %>% 
  mutate(dyad = str_c(source, " -> ", target)) 

for (i in 2001:2012) {
  print(df %>% 
  mutate(diff_w = (diff * 1e6 / target_gdp)) %>% 
  filter(year == i) %>% 
  drop_na() %>% 
  filter(rank(-diff_w) <= 10) %>% 
  ggplot(aes(x = fct_reorder(dyad, diff_w), y = diff_w)) +
  geom_point() + 
  facet_wrap(~ year, scales = "free") +
  coord_flip() +
  scale_y_continuous(labels = scales::percent) +
  labs(x = NULL, y = "Size of mismatch", caption = "(% of target country's GDP)")
  )
}

DOUBLE CHECK THIS

Top offenders by year (weighted by source country’s GDP)

for (i in 2001:2012) {
  print(df %>% 
  mutate(diff_w = (diff * 1e6 / source_gdp)) %>% 
  filter(year == i) %>% 
  drop_na() %>% 
  filter(rank(-diff_w) <= 10) %>% 
  ggplot(aes(x = fct_reorder(dyad, diff_w), y = diff_w)) +
  geom_point() + 
  facet_wrap(~ year, scales = "free") +
  coord_flip() +
  scale_y_continuous(labels = scales::percent) +
  labs(x = NULL, y = "Size of mismatch", caption = "(% of source country's GDP)")
  )
}

DOUBLE CHECK THIS

Traditional measurement of tax havens

\[ ff_i = \frac{\text{in-degree}_i}{\text{GDP}_i} \]

for (i in 2001:2012) {
  print(target_info %>% 
  mutate(ff = flow * 1e6 / target_gdp) %>% 
  group_by(target, year) %>% 
  summarise(m = sum(ff)) %>% 
  filter(year == i) %>% 
  ungroup() %>% 
  drop_na() %>% 
  filter(rank(-abs(m)) <= 10) %>% 
  ggplot(aes(x = reorder(target, m), y = m)) +
  geom_point() + 
  facet_wrap(~ year, scales = "free") +
  coord_flip() +
  scale_y_continuous(labels = scales::percent) +
  labs(x = NULL, y = "In-degree / GDP")
  )
}